You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
desispec unit tests starting failing at NERSC last Friday night August 9 for reasons related to $DESI_SPECTRO_DARK and $DESI_SPECTRO_CALIB, which might be due to PRs #2313 (more FIBERSTATUS bits) or PR #2314 (readnoise per CCD row) which were merged around that time. py/desispec/test/test_io_fibermap.py fails both when run individually and when run as part of the full test suite, albeit with different failures, which is usually a symptom of one test changing an environment variable but not resetting it, which then impacts a different test.
Note that it appears that something set $DESI_SPECTRO_CALIB=$HOME/preproc_unit_test which is then mucking up the calibfinder as part of the fibermap tests.
This is likely a problem in the tests triggered by the PR updates, not the underlying PR algorithms, but we should clean it up so that tests pass again.
The text was updated successfully, but these errors were encountered:
self.calibdir = os.path.join(os.environ['HOME'], 'preproc_unit_test') is used in test_preproc.py and test_calibfinder.py, but both of those were introduced in January of 2019. I see no other reference to 'preproc_unit_test'.
Perhaps it is not that the path is set but rather something is now expected to be at that path that isn't?
desispec unit tests starting failing at NERSC last Friday night August 9 for reasons related to $DESI_SPECTRO_DARK and $DESI_SPECTRO_CALIB, which might be due to PRs #2313 (more FIBERSTATUS bits) or PR #2314 (readnoise per CCD row) which were merged around that time. py/desispec/test/test_io_fibermap.py fails both when run individually and when run as part of the full test suite, albeit with different failures, which is usually a symptom of one test changing an environment variable but not resetting it, which then impacts a different test.
When run individually:
When run as part of the full test suite
Note that it appears that something set $DESI_SPECTRO_CALIB=$HOME/preproc_unit_test which is then mucking up the calibfinder as part of the fibermap tests.
This is likely a problem in the tests triggered by the PR updates, not the underlying PR algorithms, but we should clean it up so that tests pass again.
The text was updated successfully, but these errors were encountered: